2 * linux/fs/proc/array.c
4 * Copyright (C) 1992 by Linus Torvalds
5 * based on ideas by Darren Senn
8 * Michael. K. Johnson: stat,statm extensions.
9 * <johnsonm@stolaf.edu>
11 * Pauline Middelink : Made cmdline,envline only break at '\0's, to
12 * make sure SET_PROCTITLE works. Also removed
13 * bad '!' which forced address recalculation for
14 * EVERY character on the current page.
15 * <middelin@polyware.iaf.nl>
17 * Danny ter Haar : added cpuinfo
20 * Alessandro Rubini : profile extension.
21 * <rubini@ipvvis.unipv.it>
23 * Jeff Tranter : added BogoMips field to cpuinfo
24 * <Jeff_Tranter@Mitel.COM>
26 * Bruno Haible : remove 4K limit for the maps file
27 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
29 * Yves Arrouye : remove removal of trailing spaces in get_array.
30 * <Yves.Arrouye@marin.fdn.fr>
32 * Jerome Forissier : added per-CPU time information to /proc/stat
33 * and /proc/<pid>/cpu extension
34 * <forissier@isia.cma.fr>
35 * - Incorporation and non-SMP safe operation
36 * of forissier patch in 2.1.78 by
37 * Hans Marcus <crowbar@concepts.nl>
39 * aeb@cwi.nl : /proc/partitions
42 * Alan Cox : security fixes.
43 * <alan@lxorguk.ukuu.org.uk>
45 * Al Viro : safe handling of mm_struct
47 * Gerhard Wichert : added BIGMEM support
48 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
50 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
51 * : proc_misc.c. The rest may eventually go into
55 #include <linux/types.h>
56 #include <linux/errno.h>
57 #include <linux/time.h>
58 #include <linux/kernel.h>
59 #include <linux/kernel_stat.h>
60 #include <linux/tty.h>
61 #include <linux/string.h>
62 #include <linux/mman.h>
63 #include <linux/proc_fs.h>
64 #include <linux/ioport.h>
65 #include <linux/uaccess.h>
68 #include <linux/hugetlb.h>
69 #include <linux/pagemap.h>
70 #include <linux/swap.h>
71 #include <linux/smp.h>
72 #include <linux/signal.h>
73 #include <linux/highmem.h>
74 #include <linux/file.h>
75 #include <linux/fdtable.h>
76 #include <linux/times.h>
77 #include <linux/cpuset.h>
78 #include <linux/rcupdate.h>
79 #include <linux/delayacct.h>
80 #include <linux/seq_file.h>
81 #include <linux/pid_namespace.h>
82 #include <linux/ptrace.h>
83 #include <linux/tracehook.h>
84 #include <linux/string_helpers.h>
85 #include <linux/user_namespace.h>
87 #include <asm/pgtable.h>
88 #include <asm/processor.h>
91 static inline void task_name(struct seq_file
*m
, struct task_struct
*p
)
94 char tcomm
[sizeof(p
->comm
)];
96 get_task_comm(tcomm
, p
);
98 seq_puts(m
, "Name:\t");
99 buf
= m
->buf
+ m
->count
;
101 /* Ignore error for now */
102 buf
+= string_escape_str(tcomm
, buf
, m
->size
- m
->count
,
103 ESCAPE_SPACE
| ESCAPE_SPECIAL
, "\n\\");
105 m
->count
= buf
- m
->buf
;
110 * The task state array is a strange "bitmap" of
111 * reasons to sleep. Thus "running" is zero, and
112 * you can test for combinations of others with
115 static const char * const task_state_array
[] = {
116 "R (running)", /* 0 */
117 "S (sleeping)", /* 1 */
118 "D (disk sleep)", /* 2 */
119 "T (stopped)", /* 4 */
120 "t (tracing stop)", /* 8 */
122 "Z (zombie)", /* 32 */
125 static inline const char *get_task_state(struct task_struct
*tsk
)
127 unsigned int state
= (tsk
->state
| tsk
->exit_state
) & TASK_REPORT
;
129 BUILD_BUG_ON(1 + ilog2(TASK_REPORT
) != ARRAY_SIZE(task_state_array
)-1);
131 return task_state_array
[fls(state
)];
134 static inline void task_state(struct seq_file
*m
, struct pid_namespace
*ns
,
135 struct pid
*pid
, struct task_struct
*p
)
137 struct user_namespace
*user_ns
= seq_user_ns(m
);
138 struct group_info
*group_info
;
140 struct task_struct
*tracer
;
141 const struct cred
*cred
;
142 pid_t ppid
, tpid
= 0, tgid
, ngid
;
143 unsigned int max_fds
= 0;
146 ppid
= pid_alive(p
) ?
147 task_tgid_nr_ns(rcu_dereference(p
->real_parent
), ns
) : 0;
149 tracer
= ptrace_parent(p
);
151 tpid
= task_pid_nr_ns(tracer
, ns
);
153 tgid
= task_tgid_nr_ns(p
, ns
);
154 ngid
= task_numa_group_id(p
);
155 cred
= get_task_cred(p
);
159 max_fds
= files_fdtable(p
->files
)->max_fds
;
170 "Uid:\t%d\t%d\t%d\t%d\n"
171 "Gid:\t%d\t%d\t%d\t%d\n"
172 "FDSize:\t%d\nGroups:\t",
174 tgid
, ngid
, pid_nr_ns(pid
, ns
), ppid
, tpid
,
175 from_kuid_munged(user_ns
, cred
->uid
),
176 from_kuid_munged(user_ns
, cred
->euid
),
177 from_kuid_munged(user_ns
, cred
->suid
),
178 from_kuid_munged(user_ns
, cred
->fsuid
),
179 from_kgid_munged(user_ns
, cred
->gid
),
180 from_kgid_munged(user_ns
, cred
->egid
),
181 from_kgid_munged(user_ns
, cred
->sgid
),
182 from_kgid_munged(user_ns
, cred
->fsgid
),
185 group_info
= cred
->group_info
;
186 for (g
= 0; g
< group_info
->ngroups
; g
++)
188 from_kgid_munged(user_ns
, GROUP_AT(group_info
, g
)));
192 seq_puts(m
, "\nNStgid:");
193 for (g
= ns
->level
; g
<= pid
->level
; g
++)
194 seq_printf(m
, "\t%d",
195 task_tgid_nr_ns(p
, pid
->numbers
[g
].ns
));
196 seq_puts(m
, "\nNSpid:");
197 for (g
= ns
->level
; g
<= pid
->level
; g
++)
198 seq_printf(m
, "\t%d",
199 task_pid_nr_ns(p
, pid
->numbers
[g
].ns
));
200 seq_puts(m
, "\nNSpgid:");
201 for (g
= ns
->level
; g
<= pid
->level
; g
++)
202 seq_printf(m
, "\t%d",
203 task_pgrp_nr_ns(p
, pid
->numbers
[g
].ns
));
204 seq_puts(m
, "\nNSsid:");
205 for (g
= ns
->level
; g
<= pid
->level
; g
++)
206 seq_printf(m
, "\t%d",
207 task_session_nr_ns(p
, pid
->numbers
[g
].ns
));
212 void render_sigset_t(struct seq_file
*m
, const char *header
,
224 if (sigismember(set
, i
+1)) x
|= 1;
225 if (sigismember(set
, i
+2)) x
|= 2;
226 if (sigismember(set
, i
+3)) x
|= 4;
227 if (sigismember(set
, i
+4)) x
|= 8;
228 seq_printf(m
, "%x", x
);
234 static void collect_sigign_sigcatch(struct task_struct
*p
, sigset_t
*ign
,
237 struct k_sigaction
*k
;
240 k
= p
->sighand
->action
;
241 for (i
= 1; i
<= _NSIG
; ++i
, ++k
) {
242 if (k
->sa
.sa_handler
== SIG_IGN
)
244 else if (k
->sa
.sa_handler
!= SIG_DFL
)
249 static inline void task_sig(struct seq_file
*m
, struct task_struct
*p
)
252 sigset_t pending
, shpending
, blocked
, ignored
, caught
;
254 unsigned long qsize
= 0;
255 unsigned long qlim
= 0;
257 sigemptyset(&pending
);
258 sigemptyset(&shpending
);
259 sigemptyset(&blocked
);
260 sigemptyset(&ignored
);
261 sigemptyset(&caught
);
263 if (lock_task_sighand(p
, &flags
)) {
264 pending
= p
->pending
.signal
;
265 shpending
= p
->signal
->shared_pending
.signal
;
266 blocked
= p
->blocked
;
267 collect_sigign_sigcatch(p
, &ignored
, &caught
);
268 num_threads
= get_nr_threads(p
);
269 rcu_read_lock(); /* FIXME: is this correct? */
270 qsize
= atomic_read(&__task_cred(p
)->user
->sigpending
);
272 qlim
= task_rlimit(p
, RLIMIT_SIGPENDING
);
273 unlock_task_sighand(p
, &flags
);
276 seq_printf(m
, "Threads:\t%d\n", num_threads
);
277 seq_printf(m
, "SigQ:\t%lu/%lu\n", qsize
, qlim
);
279 /* render them all */
280 render_sigset_t(m
, "SigPnd:\t", &pending
);
281 render_sigset_t(m
, "ShdPnd:\t", &shpending
);
282 render_sigset_t(m
, "SigBlk:\t", &blocked
);
283 render_sigset_t(m
, "SigIgn:\t", &ignored
);
284 render_sigset_t(m
, "SigCgt:\t", &caught
);
287 static void render_cap_t(struct seq_file
*m
, const char *header
,
293 CAP_FOR_EACH_U32(__capi
) {
294 seq_printf(m
, "%08x",
295 a
->cap
[CAP_LAST_U32
- __capi
]);
300 static inline void task_cap(struct seq_file
*m
, struct task_struct
*p
)
302 const struct cred
*cred
;
303 kernel_cap_t cap_inheritable
, cap_permitted
, cap_effective
, cap_bset
;
306 cred
= __task_cred(p
);
307 cap_inheritable
= cred
->cap_inheritable
;
308 cap_permitted
= cred
->cap_permitted
;
309 cap_effective
= cred
->cap_effective
;
310 cap_bset
= cred
->cap_bset
;
313 render_cap_t(m
, "CapInh:\t", &cap_inheritable
);
314 render_cap_t(m
, "CapPrm:\t", &cap_permitted
);
315 render_cap_t(m
, "CapEff:\t", &cap_effective
);
316 render_cap_t(m
, "CapBnd:\t", &cap_bset
);
319 static inline void task_seccomp(struct seq_file
*m
, struct task_struct
*p
)
321 #ifdef CONFIG_SECCOMP
322 seq_printf(m
, "Seccomp:\t%d\n", p
->seccomp
.mode
);
326 static inline void task_context_switch_counts(struct seq_file
*m
,
327 struct task_struct
*p
)
329 seq_printf(m
, "voluntary_ctxt_switches:\t%lu\n"
330 "nonvoluntary_ctxt_switches:\t%lu\n",
335 static void task_cpus_allowed(struct seq_file
*m
, struct task_struct
*task
)
337 seq_printf(m
, "Cpus_allowed:\t%*pb\n",
338 cpumask_pr_args(&task
->cpus_allowed
));
339 seq_printf(m
, "Cpus_allowed_list:\t%*pbl\n",
340 cpumask_pr_args(&task
->cpus_allowed
));
343 int proc_pid_status(struct seq_file
*m
, struct pid_namespace
*ns
,
344 struct pid
*pid
, struct task_struct
*task
)
346 struct mm_struct
*mm
= get_task_mm(task
);
349 task_state(m
, ns
, pid
, task
);
357 task_seccomp(m
, task
);
358 task_cpus_allowed(m
, task
);
359 cpuset_task_status_allowed(m
, task
);
360 task_context_switch_counts(m
, task
);
364 static int do_task_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
365 struct pid
*pid
, struct task_struct
*task
, int whole
)
367 unsigned long vsize
, eip
, esp
, wchan
= 0;
369 int tty_pgrp
= -1, tty_nr
= 0;
370 sigset_t sigign
, sigcatch
;
372 pid_t ppid
= 0, pgid
= -1, sid
= -1;
375 struct mm_struct
*mm
;
376 unsigned long long start_time
;
377 unsigned long cmin_flt
= 0, cmaj_flt
= 0;
378 unsigned long min_flt
= 0, maj_flt
= 0;
379 cputime_t cutime
, cstime
, utime
, stime
;
380 cputime_t cgtime
, gtime
;
381 unsigned long rsslim
= 0;
382 char tcomm
[sizeof(task
->comm
)];
385 state
= *get_task_state(task
);
386 vsize
= eip
= esp
= 0;
387 permitted
= ptrace_may_access(task
, PTRACE_MODE_READ_FSCREDS
| PTRACE_MODE_NOAUDIT
);
388 mm
= get_task_mm(task
);
390 vsize
= task_vsize(mm
);
392 eip
= KSTK_EIP(task
);
393 esp
= KSTK_ESP(task
);
397 get_task_comm(tcomm
, task
);
399 sigemptyset(&sigign
);
400 sigemptyset(&sigcatch
);
401 cutime
= cstime
= utime
= stime
= 0;
404 if (lock_task_sighand(task
, &flags
)) {
405 struct signal_struct
*sig
= task
->signal
;
408 struct pid
*pgrp
= tty_get_pgrp(sig
->tty
);
409 tty_pgrp
= pid_nr_ns(pgrp
, ns
);
411 tty_nr
= new_encode_dev(tty_devnum(sig
->tty
));
414 num_threads
= get_nr_threads(task
);
415 collect_sigign_sigcatch(task
, &sigign
, &sigcatch
);
417 cmin_flt
= sig
->cmin_flt
;
418 cmaj_flt
= sig
->cmaj_flt
;
419 cutime
= sig
->cutime
;
420 cstime
= sig
->cstime
;
421 cgtime
= sig
->cgtime
;
422 rsslim
= ACCESS_ONCE(sig
->rlim
[RLIMIT_RSS
].rlim_cur
);
424 /* add up live thread stats at the group level */
426 struct task_struct
*t
= task
;
428 min_flt
+= t
->min_flt
;
429 maj_flt
+= t
->maj_flt
;
430 gtime
+= task_gtime(t
);
431 } while_each_thread(task
, t
);
433 min_flt
+= sig
->min_flt
;
434 maj_flt
+= sig
->maj_flt
;
435 thread_group_cputime_adjusted(task
, &utime
, &stime
);
439 sid
= task_session_nr_ns(task
, ns
);
440 ppid
= task_tgid_nr_ns(task
->real_parent
, ns
);
441 pgid
= task_pgrp_nr_ns(task
, ns
);
443 unlock_task_sighand(task
, &flags
);
446 if (permitted
&& (!whole
|| num_threads
< 2))
447 wchan
= get_wchan(task
);
449 min_flt
= task
->min_flt
;
450 maj_flt
= task
->maj_flt
;
451 task_cputime_adjusted(task
, &utime
, &stime
);
452 gtime
= task_gtime(task
);
455 /* scale priority and nice values from timeslices to -20..20 */
456 /* to make it look like a "normal" Unix priority/nice value */
457 priority
= task_prio(task
);
458 nice
= task_nice(task
);
460 /* convert nsec -> ticks */
461 start_time
= nsec_to_clock_t(task
->real_start_time
);
463 seq_printf(m
, "%d (%s) %c", pid_nr_ns(pid
, ns
), tcomm
, state
);
464 seq_put_decimal_ll(m
, ' ', ppid
);
465 seq_put_decimal_ll(m
, ' ', pgid
);
466 seq_put_decimal_ll(m
, ' ', sid
);
467 seq_put_decimal_ll(m
, ' ', tty_nr
);
468 seq_put_decimal_ll(m
, ' ', tty_pgrp
);
469 seq_put_decimal_ull(m
, ' ', task
->flags
);
470 seq_put_decimal_ull(m
, ' ', min_flt
);
471 seq_put_decimal_ull(m
, ' ', cmin_flt
);
472 seq_put_decimal_ull(m
, ' ', maj_flt
);
473 seq_put_decimal_ull(m
, ' ', cmaj_flt
);
474 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(utime
));
475 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(stime
));
476 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cutime
));
477 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cstime
));
478 seq_put_decimal_ll(m
, ' ', priority
);
479 seq_put_decimal_ll(m
, ' ', nice
);
480 seq_put_decimal_ll(m
, ' ', num_threads
);
481 seq_put_decimal_ull(m
, ' ', 0);
482 seq_put_decimal_ull(m
, ' ', start_time
);
483 seq_put_decimal_ull(m
, ' ', vsize
);
484 seq_put_decimal_ull(m
, ' ', mm
? get_mm_rss(mm
) : 0);
485 seq_put_decimal_ull(m
, ' ', rsslim
);
486 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->start_code
: 1) : 0);
487 seq_put_decimal_ull(m
, ' ', mm
? (permitted
? mm
->end_code
: 1) : 0);
488 seq_put_decimal_ull(m
, ' ', (permitted
&& mm
) ? mm
->start_stack
: 0);
489 seq_put_decimal_ull(m
, ' ', esp
);
490 seq_put_decimal_ull(m
, ' ', eip
);
491 /* The signal information here is obsolete.
492 * It must be decimal for Linux 2.0 compatibility.
493 * Use /proc/#/status for real-time signals.
495 seq_put_decimal_ull(m
, ' ', task
->pending
.signal
.sig
[0] & 0x7fffffffUL
);
496 seq_put_decimal_ull(m
, ' ', task
->blocked
.sig
[0] & 0x7fffffffUL
);
497 seq_put_decimal_ull(m
, ' ', sigign
.sig
[0] & 0x7fffffffUL
);
498 seq_put_decimal_ull(m
, ' ', sigcatch
.sig
[0] & 0x7fffffffUL
);
501 * We used to output the absolute kernel address, but that's an
502 * information leak - so instead we show a 0/1 flag here, to signal
503 * to user-space whether there's a wchan field in /proc/PID/wchan.
505 * This works with older implementations of procps as well.
512 seq_put_decimal_ull(m
, ' ', 0);
513 seq_put_decimal_ull(m
, ' ', 0);
514 seq_put_decimal_ll(m
, ' ', task
->exit_signal
);
515 seq_put_decimal_ll(m
, ' ', task_cpu(task
));
516 seq_put_decimal_ull(m
, ' ', task
->rt_priority
);
517 seq_put_decimal_ull(m
, ' ', task
->policy
);
518 seq_put_decimal_ull(m
, ' ', delayacct_blkio_ticks(task
));
519 seq_put_decimal_ull(m
, ' ', cputime_to_clock_t(gtime
));
520 seq_put_decimal_ll(m
, ' ', cputime_to_clock_t(cgtime
));
522 if (mm
&& permitted
) {
523 seq_put_decimal_ull(m
, ' ', mm
->start_data
);
524 seq_put_decimal_ull(m
, ' ', mm
->end_data
);
525 seq_put_decimal_ull(m
, ' ', mm
->start_brk
);
526 seq_put_decimal_ull(m
, ' ', mm
->arg_start
);
527 seq_put_decimal_ull(m
, ' ', mm
->arg_end
);
528 seq_put_decimal_ull(m
, ' ', mm
->env_start
);
529 seq_put_decimal_ull(m
, ' ', mm
->env_end
);
531 seq_printf(m
, " 0 0 0 0 0 0 0");
534 seq_put_decimal_ll(m
, ' ', task
->exit_code
);
536 seq_put_decimal_ll(m
, ' ', 0);
544 int proc_tid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
545 struct pid
*pid
, struct task_struct
*task
)
547 return do_task_stat(m
, ns
, pid
, task
, 0);
550 int proc_tgid_stat(struct seq_file
*m
, struct pid_namespace
*ns
,
551 struct pid
*pid
, struct task_struct
*task
)
553 return do_task_stat(m
, ns
, pid
, task
, 1);
556 int proc_pid_statm(struct seq_file
*m
, struct pid_namespace
*ns
,
557 struct pid
*pid
, struct task_struct
*task
)
559 unsigned long size
= 0, resident
= 0, shared
= 0, text
= 0, data
= 0;
560 struct mm_struct
*mm
= get_task_mm(task
);
563 size
= task_statm(mm
, &shared
, &text
, &data
, &resident
);
567 * For quick read, open code by putting numbers directly
569 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
570 * size, resident, shared, text, data);
572 seq_put_decimal_ull(m
, 0, size
);
573 seq_put_decimal_ull(m
, ' ', resident
);
574 seq_put_decimal_ull(m
, ' ', shared
);
575 seq_put_decimal_ull(m
, ' ', text
);
576 seq_put_decimal_ull(m
, ' ', 0);
577 seq_put_decimal_ull(m
, ' ', data
);
578 seq_put_decimal_ull(m
, ' ', 0);
584 #ifdef CONFIG_CHECKPOINT_RESTORE
586 get_children_pid(struct inode
*inode
, struct pid
*pid_prev
, loff_t pos
)
588 struct task_struct
*start
, *task
;
589 struct pid
*pid
= NULL
;
591 read_lock(&tasklist_lock
);
593 start
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
598 * Lets try to continue searching first, this gives
599 * us significant speedup on children-rich processes.
602 task
= pid_task(pid_prev
, PIDTYPE_PID
);
603 if (task
&& task
->real_parent
== start
&&
604 !(list_empty(&task
->sibling
))) {
605 if (list_is_last(&task
->sibling
, &start
->children
))
607 task
= list_first_entry(&task
->sibling
,
608 struct task_struct
, sibling
);
609 pid
= get_pid(task_pid(task
));
617 * We might miss some children here if children
618 * are exited while we were not holding the lock,
619 * but it was never promised to be accurate that
622 * "Just suppose that the parent sleeps, but N children
623 * exit after we printed their tids. Now the slow paths
624 * skips N extra children, we miss N tasks." (c)
626 * So one need to stop or freeze the leader and all
627 * its children to get a precise result.
629 list_for_each_entry(task
, &start
->children
, sibling
) {
631 pid
= get_pid(task_pid(task
));
637 read_unlock(&tasklist_lock
);
641 static int children_seq_show(struct seq_file
*seq
, void *v
)
643 struct inode
*inode
= seq
->private;
646 pid
= pid_nr_ns(v
, inode
->i_sb
->s_fs_info
);
647 seq_printf(seq
, "%d ", pid
);
652 static void *children_seq_start(struct seq_file
*seq
, loff_t
*pos
)
654 return get_children_pid(seq
->private, NULL
, *pos
);
657 static void *children_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
661 pid
= get_children_pid(seq
->private, v
, *pos
+ 1);
668 static void children_seq_stop(struct seq_file
*seq
, void *v
)
673 static const struct seq_operations children_seq_ops
= {
674 .start
= children_seq_start
,
675 .next
= children_seq_next
,
676 .stop
= children_seq_stop
,
677 .show
= children_seq_show
,
680 static int children_seq_open(struct inode
*inode
, struct file
*file
)
685 ret
= seq_open(file
, &children_seq_ops
);
689 m
= file
->private_data
;
695 int children_seq_release(struct inode
*inode
, struct file
*file
)
697 seq_release(inode
, file
);
701 const struct file_operations proc_tid_children_operations
= {
702 .open
= children_seq_open
,
705 .release
= children_seq_release
,
707 #endif /* CONFIG_CHECKPOINT_RESTORE */